Ultrasonic imaging is performed by constructing spatial coherence images of a target having microbubbles in it. The basis for this approach is the observation that the spatial coherence of microbubbles differs from the spatial coherence of tissue and the spatial coherence of image noise. Therefore, imaging based on spatial coherence provides a way to suppress noise signals and tissue signals relative to the microbubble signals.
|
1. A method for ultrasound medical imaging, the method comprising:
providing a contrast agent to a target being imaged, wherein the contrast agent includes microbubbles;
providing incident acoustic radiation to the target with an ultrasound transducer array, wherein the target provides an acoustic signal responsive to the incident acoustic radiation;
receiving the acoustic signal with the ultrasound transducer array;
providing an ultrasound image of the target by applying a beamforming method to the acoustic signal;
wherein the beamforming method includes:
i) estimating a spatial coherence function of the target;
ii) integrating the spatial coherence function over a predetermined integration window;
wherein some or all of the microbubbles are bound to the target;
wherein the ultrasound image of the target provides improved sensitivity to bound microbubbles relative to noise in the ultrasound image.
2. The method of
3. The method of
4. The method of
5. The method of
6. The method of
wherein a first part of an imaging sequence has an intensity of the incident acoustic radiation that is sufficiently low that the microbubbles do not burst during the first part, and
wherein a second part of the imaging sequence has an intensity of the incident acoustic radiation that is sufficient to burst the microbubbles;
wherein a difference image between a pre-burst image and a post-burst image is formed to enhance microbubble features in the difference image.
7. The method of
8. The method of
9. The method of
10. The method of
11. The method of
12. The method of
13. The method of
14. The method of
|
This application claims the benefit of U.S. provisional patent application 62/317,122, filed on Apr. 1, 2016, and hereby incorporated by reference in its entirety.
This invention was made with Government support under contract EB013661 awarded by the National Institutes of Health. The Government has certain rights in the invention.
This invention relates to ultrasound imaging.
Ultrasound imaging has been employed for various medical applications. Methods of improving imaging performance in such cases are of general interest. One method of improving ultrasound imaging is the use of microbubbles as contrast agents. For example, microbubbles can be configured to specifically bind to tumor antigens to aid in cancer detection. However, new problems can arise in connection with the use of microbubbles as contrast agents. In particular, if it is desired to have the microbubbles remain intact during imaging (as opposed to the more conventional approach of bursting the microbubbles when imaging), then it is necessary to have a relatively low incident acoustic intensity, which undesirably tends to reduce image quality. Accordingly, it would be an advance in the art to provide improved acoustic imaging using microbubbles as contrast agents.
In this work, ultrasonic imaging is performed by constructing spatial coherence images of a target having microbubbles in it, as opposed to the more conventional signal magnitude images. The basis for this approach is our observation that the spatial coherence of microbubbles differs from the spatial coherence of tissue and the spatial coherence of image noise (see
Section A describes general principles relating to embodiments of the invention, and section B is a detailed experimental example.
Image beamforming to provide an ultrasound image of the target is performed in processor 118. In particular, this beamforming includes at least the following two steps:
The spatial coherence function is a measure of the correlation function between signals with a given spacing, or lag, of m elements. This function is defined at every field point x in the ultrasound image. Thus Rm(x) is the measured correlation coefficient of transducer element signals from field point x for transducer elements having a relative lag of m. For two-dimensional transducer arrays, m refers to a two-dimensional lag, with components in the two dimensions of the array. In some cases, these measured correlation coefficients are computed by averaging over a correlation window (e.g., a one wavelength axial signal window centered at x). In other cases no averaging is needed to provide these correlation coefficients, leading to so-called ‘single-pixel’ computations. Here field point x can be a pair of two coordinates (x=(x1, x2)) or a triplet of three coordinates (x=x1, x2, x3) for 2D and 3D cases, respectively.
Integration of a spatial coherence function can be performed by analog integration over the appropriate range, but in practice this integration is typically performed by a discrete summation over an appropriate range of indices.
Preferably the microbubbles are configured to have a spatial coherence distinct from a spatial coherence of tissue in the target and distinct from a spatial coherence of measurement noise.
For 1-D transducer arrays, the ultrasound transducer array can have N elements and the predetermined integration window is preferably a lag range from 1 to N/2 and is more preferably a lag range from 1 to N/4. For 2-D transducer arrays, the ultrasound transducer array can have Nx×Ny elements and the predetermined integration window is preferably a lag range from 1 to Nx/2 and from 1 to Ny/2 and is more preferably a lag range from 1 to Nx/4 and from 1 to Ny/4 and is still more preferably the set of all 2-D lags (mx, my) such that (1−mx/Nx) (1−my/Ny)>0.75. Further details on suitable beamforming methods for short-lag spatial coherence imaging are given in an article by Hyun et al. (IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, v61n7, pp. 1101-1112, 2014) hereby incorporated by reference in its entirety.
In some embodiments, the intensity of the incident acoustic radiation is sufficiently low that the microbubbles do not burst during imaging. In other embodiments imaging is divided into two parts where the first part has an incident acoustic intensity that is sufficiently low that the microbubbles do not burst, and where the second part has an incident acoustic intensity sufficient to burst the microbubbles. Bursting the microbubbles can be used to provide increased acoustic signal and/or to deliver a therapeutic agent.
The incident acoustic radiation is preferably configured as pulse sequences having linear echoes that substantially cancel in the acoustic signal. The microbubbles are preferably configured to provide a nonlinear acoustic signal at one or more frequencies distinct from frequencies in the incident acoustic radiation.
The microbubbles can be configured to specifically bind to features within the target.
Contrast-enhanced ultrasonography (CEUS) utilizes microbubble (MB) contrast agents for enhanced anatomical, functional, and molecular imaging. MB contrast agents have been used effectively in obtaining better images of the heart, liver, kidney, and breast vasculature, and have been used to improve ultrasound imaging in a variety of other applications. MBs typically include a biocompatible shell with a gas core, ranging from 1 to 6 microns in diameter. Despite their small size, MB contrast agents have a strong scattering response to insonification, making them easily detectable with ultrasound. MBs also have a harmonic acoustic signature that is distinct from that of tissue parenchyma. These differences can be leveraged to isolate the MB signal via frequency-based techniques such as harmonic imaging or acoustic angiography.
MBs can also be coated with ligands and antibodies to target specific biomarkers for molecular CEUS (MCEUS) imaging applications. For example, cancerous tissue can be identified by targeting MBs to antigens that are often over-expressed by growing tumors, such as vascular endothelial growth factor receptor 2 (VEGFR2). By combining the sensitivity of ultrasound to MBs and the specificity of MBs to the tumor antigens, MCEUS can be used to aid in the early detection of cancer. Several preliminary studies in preclinical applications have shown the potential for MCEUS.
However, there are numerous challenges towards translating MCEUS to clinical use. MBs are fragile and are easily destroyed by the pressures used in diagnostic ultrasound imaging, and require specially designed low pressure pulse sequences with a low mechanical index (MI), ranging from 0.1 to 0.5. The low pressures required to keep the microbubbles intact result in significantly degraded signal-to-noise ratio (SNR) of the received echoes. This effect is compounded by the presence of subcutaneous tissue layers between the transducer and the tumor, an obstacle that is often ignored in preclinical exams of superficial tumors. The tissue may generate reverberation clutter and phase aberration artifacts, in addition to frequency-dependent attenuation of high frequency echoes.
Much of the current research efforts are centered on developing novel pulse sequences and techniques to improve binding of MBs to receptors. These techniques are used together with the classical delay-and-sum (DAS) beamformer to form images based on the magnitude of the echo. Despite its ubiquity, the DAS beamformer is unsophisticated, and in low SNR imaging environments, is especially subject to imaging artifacts caused by thermal noise and clutter. More sensitive and robust imaging techniques are needed as MCEUS research transitions into clinical applications, where imaging conditions are not ideal.
Short-lag spatial coherence (SLSC) is an alternative beamforming technique that makes images of the spatial coherence of the echo, rather than of its magnitude. The spatial coherence of an echo is a measure of how similar the wavefront is between adjacent elements. In theory, signals from diffuse scatterers such as tissue should be partially coherent, while those from incoherent noises like reverberation clutter should be incoherent. SLSC imaging has demonstrated improved lesion detectability in simulations and in vivo. Higher harmonics of the echoes demonstrate similar trends, and have yielded similar results when used in conjunction with SLSC imaging. These reasons lead us to believe that SLSC is well suited for low SNR in vivo imaging. Furthermore, because beamforming is performed after data acquisition, SLSC can be applied in conjunction with any advances in other aspects of MCEUS, such as pulse sequence techniques.
In this work, we apply SLSC beamforming to MCEUS in a flow channel phantom experiment as well as in a mouse model of hepatocellular carcinoma, comparing the results against those of the conventional DAS beamformer.
B2a) Beamforming
Conventional CEUS and SLSC-CEUS images were reconstructed using the DAS and SLSC beamformers, respectively. Let si[n] correspond to the n-th sample of the focused complex signal from element i, reconstructed by applying the proper focal delays. For an aperture of N elements, the output of the DAS beamformer is computed as:
IDAS[n]=|Σi=1Nsi[n]| (1)
The SLSC beamformer includes two steps: estimating the spatial coherence function, and integrating the coherence function at short lags. The spatial coherence function is a measure of the average correlation coefficient between signals with a given spacing, or lag, of m elements. On a 1D array, this can be computed as
where ri,i+m[n] is the correlation coefficient between elements i and i+m:
Historically, the correlation coefficient was computed over a kernel of length T, as in Eq. 3, typically chosen as a one wavelength axial signal window centered at n. Here, we use a kernel-less approach, using a single-sample kernel:
The output of the SLSC beamformer is then computed by integrating the spatial coherence function over “short” lags, i.e. for m≤M:
ISLSC[n]=Σm=1M={circumflex over (R)}m[n]. (5)
The threshold M is usually set to be 25% of the aperture M=N/4.
B2b) Microbubble Imaging System
A Verasonics (Verasonics, Redmond, Wash.) Vantage 256 research scanner was used to acquire the raw channel data from 128 elements of an L12-3v transducer in real time. The harmonic component of the signal was obtained with a CPS approach, combined with a plane wave synthetic aperture technique. Throughout these experiments, three plane waves were transmitted at each of 7 angles over a span of 18° for a total of 21 pulses per image frame. Each triplet of pulses included a positive, negative, and positive polarity two cycle transmission at 4.5 MHz, where the positive and negative pulses were 180° out of phase. The received echoes were bandpass filtered at the second harmonic frequency (9 MHz) to detect the non-linear MB response. Each triplet of signals was summed together to cancel out the linear component, further emphasizing the harmonic signals.
The data was then transferred to a host computer for processing. Both DAS and SLSC beamformers were implemented as a MATLAB® (Mathworks, Natick, Mass.) MEX function, written in C++ with the CUDA (Compute Unified Device Architecture) application programming interface (API) from NVIDIA (NVIDIA, Santa Clara, Calif.). An NVIDIA Quadro K620 graphics processing unit (GPU) was used to focus, beamform, and display side-by-side conventional CEUS and SLSC-CEUS images in real time at >20 frames per second (fps).
B2c) Experimental Methods
The in vitro experiments were performed in an agarose flow channel phantom. The phantom was made with an agar gel (2% weight-by-volume) mixed with graphite (2%) for tissue-mimicking scattering. The gel was poured into a container lined with ρ-c rubber and with a removable 2 mm diameter plastic tube spanning from wall to wall across the center. The tube was removed after the gel had cooled and set, leaving behind a wall-less flow channel. A solution containing clinical grade BR55 (Bracco, Milan) was injected into the flow channel with a syringe. BR55 is a MB contrast agent targeted to VEGFR2. In this study, BR55 was used as a non-targeted MB because the phantom did not contain any VEGFR2 receptors. Concentrations of 5×104, 5×105, 5×106, and 5×107 MB/mL were used in this preliminary study. The lowest and highest concentrations were further tested in high-noise environments by reducing the transmit voltage by a factor of 3.
In vivo experiments were performed in a mouse model of hepatocellular carcinoma with a xenografted subcutaneous tumor on the hind limb. A 1 cm thick layer of degassed porcine tissue was placed between the transducer and tumor to mimic clinical imaging conditions.
Idiff[n]=Ipre[n]−Ipost[n]. (6)
To enhance the image, multiple frames were averaged together for both Ipre[n] and Ipost[n], with simple motion correction applied using the MATLAB Image Processing Toolbox.
In the experiments, the MB sensitivity of each beamformer was measured as follows:
where RMS is the root-mean-square value, I is either the conventional CEUS or SLSC-CEUS image, and nMB and nTissue correspond to samples in regions of interest (ROIs) of MB signal and tissue signal, respectively.
B3a) Phantom Experiments
The MBs were easily visible in the flow channel phantom for all MB concentrations. The measured SNRs are reported in Table 1. In all cases, the SNR was very high for both conventional CEUS and SLSC-CEUS, indicating that the MBs were easily detected. In each case, SLSC-CEUS had slightly higher SNR, with the greatest disparity at the lowest concentration of MBs tested.
TABLE 1
Imaging SNR in Agarose Phantom
Concentration
CEUS SNR
SLSC-CEUS SNR
5 × 107 MB/mL
18.6 dB
20.9 dB
5 × 106 MB/mL
23.9 dB
24.0 dB
5 × 105 MB/mL
19.5 dB
20.5 dB
5 × 104 MB/mL
16.5 dB
22.0 dB
In the high-noise imaging environment, a greater difference between the two methods was observed, as shown in Table 2, with SLSC-CEUS outperforming CEUS.
TABLE 2
Imaging SNR in Agarose Phantom in Noisy Conditions
Concentration
CEUS SNR
SLSC-CEUS SNR
5 × 107 MB/mL
17.0 dB
21.4 dB
5 × 104 MB/mL
8.5 dB
15.0 dB
B3b) In Vivo Experiments
Pre-burst, post-burst, and difference images are shown in
Similarly,
The sensitivities of the SLSC-CEUS images are plotted against those of the conventional CEUS images in
In this study, in vivo imaging conditions were mimicked using a layer of porcine tissue. SLSC-CEUS improved the sensitivity in every acquisition, with an average increase of 65% in SNR. The clutter and noise generated in vivo is often spatially incoherent but high in magnitude. The conventional CEUS beamformer, which detects magnitude, interprets the noise as signal. The SLSC beamformer instead detects regions that scatter off echoes with high spatial coherence, regardless of magnitude. This allows SLSC-CEUS to identify echo sources that are weak in magnitude but high in spatial coherence, such as individually bound MBs, and to suppress noise that is high in magnitude and spatially incoherent.
The SLSC beamformer improves sensitivity to targeted MBs in conditions similar to clinical imaging by utilizing the spatial coherence of the echo, and may pave the way for the early detection of cancer in humans.
Dahl, Jeremy Joseph, Hyun, Dongwoon, Willmann, Juergen K.
Patent | Priority | Assignee | Title |
Patent | Priority | Assignee | Title |
5540909, | Sep 28 1994 | TARGESON, INC ; TARGESON INC | Harmonic ultrasound imaging with microbubbles |
6951540, | May 10 2002 | Regents of the University of Minnesota | Ultrasound imaging system and method using non-linear post-beamforming filter |
9237898, | Jul 21 2009 | University of Virginia Patent Foundation | Systems and methods for ultrasound imaging and insonation of microbubbles |
9329260, | Dec 16 2010 | UNIVERSITÉ PARIS CITÉ | Method and device for ultrasound imaging |
9532769, | Sep 04 2009 | SUNNYBROOK RESEARCH INSTITUTE | Systems, methods, and computer readable media for high frequency contrast imaging and image-guided therapeutics |
20040059225, | |||
20080281205, | |||
20130109971, | |||
20150023881, | |||
20160262727, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Apr 01 2016 | DAHL, JEREMY JOSEPH | The Board of Trustees of the Leland Stanford Junior University | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 041819 | /0824 | |
Apr 01 2016 | HYUN, DONGWOON | The Board of Trustees of the Leland Stanford Junior University | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 041819 | /0824 | |
Apr 01 2016 | WILLMANN, JUERGEN K | The Board of Trustees of the Leland Stanford Junior University | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 041819 | /0824 | |
Mar 30 2017 | The Board of Trustees of the Leland Stanford Junior University | (assignment on the face of the patent) | / |
Date | Maintenance Fee Events |
Apr 15 2024 | REM: Maintenance Fee Reminder Mailed. |
Date | Maintenance Schedule |
Aug 25 2023 | 4 years fee payment window open |
Feb 25 2024 | 6 months grace period start (w surcharge) |
Aug 25 2024 | patent expiry (for year 4) |
Aug 25 2026 | 2 years to revive unintentionally abandoned end. (for year 4) |
Aug 25 2027 | 8 years fee payment window open |
Feb 25 2028 | 6 months grace period start (w surcharge) |
Aug 25 2028 | patent expiry (for year 8) |
Aug 25 2030 | 2 years to revive unintentionally abandoned end. (for year 8) |
Aug 25 2031 | 12 years fee payment window open |
Feb 25 2032 | 6 months grace period start (w surcharge) |
Aug 25 2032 | patent expiry (for year 12) |
Aug 25 2034 | 2 years to revive unintentionally abandoned end. (for year 12) |